• No results found

T Expert studies system

I: AN EXPERT ADVISORY SYSTEM FOR AIDS/H1V

1.5 The Information System Approach 1 Introduction

1.5.3.1 Knowledge Acquisition

The knowledge acquisition techniques that have been used in the past are described in the next chapter o f this thesis (section 2.3). Here, the tools that were used in the development o f the AIDS Information System are described. The two knowledge engineers in the development team had by now gained quite a lot of

medical background in the general area of AIDS/HIV. A further study o f

approximately one month had been carried out to become familiar with the terminology and concepts that would be used in the modules that would make up the AIDS Information System.

It was important to get familiar with the subject domain before experts were approached because they are under great pressure and have very little time to talk to knowledge engineers (Fox et al 1985).

Published Text

There was a great deal of textual information available on AIDS/HIV in textbooks, booklets, leaflets, and articles in medical journals. The textbooks on AIDS/HIV included Farthing's 'Colour Atlas on AIDS'; Miller's 'Living with AIDS and HIV'; Miller, Weber and Green's 'The Management o f AIDS Patients'; and Daniels’ 'AIDS Questions and Answers'. From these titles it can be seen that a wide range of books were available, catering for several user groups with different needs. One of the main problems with even the latest available books was that some o f the information contained in them had become outdated by the time that they had been

I: AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 43

published, because new information about AIDS and HIV was being produced fairly rapidly. Nevertheless, the books still provided a rich source of useful information.

The Terrence Higgins Trust, a registered charity set up to inform, advice and help on AIDS, had produced a number o f booklets and leaflets on AIDS/HIV. These and others produced by the Health Education Authority and the Department of Health were useful both in terms of the information content and the way in which the knowledge was divided into separate sections to cater for different groups.

There were articles on AIDS/HIV appearing regularly in medical journals like the Launcet, the British Medical Journal and the New England Journal o f Medicine. These were very important because they contained information on the latest research findings, new terminologies and new definitions of AIDS. These were used primarily for maintaining, refining and updating the information obtained from other textual sources. For the Guidelines module, the two main sources, St. Mary's Control o f Infection Pack (St. Mary's Hospital 1987) and guidelines from the Amalgamated Committee on Disease Pathogens (COHSE 1987), were used.

These materials provided a good source of information for developing quite a large knowledge base for use within the two modules that would make up the AIDS Information System.

Interviews with Experts

After an initial prototype system had been designed using text based materials, the ESDSU decided to show the system to collaborating experts and at the same time, have knowledge engineering sessions with them. The importance o f having an

I AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 4 4

operational system to show experts before any interviews for knowledge acquisition took place have been emphasised in the discussion following the development of the AIDS Expert System. The main experts used were Dr. Brittain and Dr. Drake from NWHA, and Dr. Farthing from St. Stephen's Hospital in London. Meetings were arranged on a regular basis with these experts, so that they could see the system developing in the light o f their comments and suggestions for improvements.

Other experts were also used although on an irregular basis, because the time that they had available was very limited. These included Dr. Pinching, a consultant in sexually transmitted diseases at St. Mary's Hospital in London; Professor Adler, a consultant in sexually transmitted diseases at Middlesex Hospital in London; and experts at Healthline. The expertise derived from Dr. Pinching, Professor Adler and Dr. Farthing was invaluable because they were recognised as the leading experts in AIDS/HIV. They were also based in the three London Hospitals which had become established as centres of excellence because of their direct experience with a large number o f AIDS/HIV patients.

All interviews between the expert and the knowledge engineer were recorded. This was important as it enabled the knowledge engineer to discuss points with the expert without slowing him down or distracting himself by continuously having to make notes on the conversations. The fact that the knowledge engineer had become familiar with the subject domain was of paramount importance, as experts find it frustrating to have to explain basic facts, terms and phrases unnecessarily. This would also have slowed the process of elicitation of knowledge from the expert and consequently, the development of the system as the time experts can allocate is already very limited (Fox et al 1985).

1: AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 45

During the development of the Information System, the experts could see their input of information/knowledge more directly. The speed of development was also an important factor. They were willing to devote more o f their time in the development, especially as they could see their efforts being translated quickly into

a computer system. The knowledge acquisition techniques used for the

development of the AIDS Expert System are described in section 1.4.

During the knowledge engineering sessions, the experts suggested the development o f two further modules which they felt were very important in the area of AIDS. The first of these was suggested by Dr Farthing, who wanted a clinical module to assist in the identification, treatment and management of diseases found in AIDS/HIV patients. The second was a counselling module and this was identified by a number of experts as well as the development team, but in particular by Dr Drake. The important need for these two modules and a brief description o f the contents o f each is given later (see section 1.5.3.2).

These two additional modules were developed along the same lines as the General Public and Health Professionals systems. Existing published material was used in order to build a solid foundation o f knowledge for the system before these were refined using formal knowledge engineering sessions with the experts. Again Dr Brittain and Dr Drake were used regularly to discuss the important topic areas on an informal basis.

Healthline operated a 24 hour 'phone-in' service which provided advice on a number of topics including AIDS. The telephone operators were not medically qualified. They would either answer the caller directly or play a short tape which was a recording o f an expert talking on the specified subject.

I AN EXPERT ADVISORY SYSTEM FOR AIDS'HIV 46

A number o f meetings were arranged with Healthline where the AIDS Information System was demonstrated during its various development stages. It was soon realised that Healthline operators could use the system as a tool to supplement their own understanding. They could also use the system to answer calls directly. Healthline's contribution to the AIDS Information System was invaluable because they were able to provide a good indication o f the types of queries that were commonly asked by the general public.

During the development of the AIDS Information System, approximately forty hours of tape recording had been made. This represented a significant amount of knowledge especially as a voice activated cassette recorder was used for the majority o f the interviews. The tapes were later played and with the help of any textual information which the experts may have pointed out for certain references, together with information from other sources which had already been acquired, the knowledge engineer structured and transformed this knowledge into the knowledge base of the AIDS Information System.

As with the development of the AIDS Expert System, an iterative process was used for building the Information System (see figure 1.2). This involved building a prototype followed by a cyclic process o f extending, refining and modifying.

Two of the main problems that were encountered during the development of the AIDS Expert System were also prominent during the construction of the knowledge base for the AIDS Information System.

The first o f these problems was that of differing advice and procedures between different experts. This was more noticeable in the clinical module where the recommended drug therapies and investigative procedures differed among the

I AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 47

main London AIDS hospitals. The development team attempted to resolve these by giving the user alternatives to choose from. This was preferred to presenting just the thoughts of a particular group and hiding the alternatives.

As already mentioned, although the time that the experts could devote to the development of the Information System was limited, one of the biggest problems encountered during the development process was the bottle-neck which occurred during knowledge acquisition. For example, this occurred during several three­ weekly periods of stay at St Stephen's Hospital when working with Dr Charles Farthing. The stays were not longer because the developers o f the system could not keep up with the knowledge that had been acquired and the changes demanded by the experts. This was frustrating for both the expert and the development team.

Feedback from users

One of the features provided by the AIDS Information System was the facility for the user to enter any comment or question, which would get stored in a text file. This 'feedback facility' could be activated by the user at any point whilst using the system. When the latest system was left with the user-groups who were evaluating the system, this facility provided a useful way o f getting feedback about the system including spelling mistakes, comments on the structure, the instructions and the ease with which it could be used.

These comments would then be viewed later by the knowledge engineer and used to refine, modify and update the system depending on the person who made the comments. Some of the comments needed to be discussed with the experts. If this text file contained suggestions on the layout or design o f the system, these

1 : AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 48

comments were passed on to the person responsible for programming in the development team.

A Conceptual Model of the System

There were two aspects of the AIDS Information System that were most important, particularly for menu based information retrieval. These were the structure o f the system in terms o f menu and sub-menu headings, and the contents of the reports. One o f the problems faced by the knowledge engineer was the fact that not all experts had access to a computer. Therefore, they could not use the system for evaluation let alone use the feedback facility described above.

One of the main experts being used for the development of the Information System was Dr. Charles Farthing. The knowledge engineers spent several three-weekly periods at St. Stephen's Hospital, knowledge engineering Dr. Farthing before returning to the University to refine and update the Information System.

A conceptual model for the menu structure o f the general public module was made to try and overcome the problem o f the expert not having access to a PC. This consisted simply o f pieces o f paper (pages) of different lengths, each with a different heading on it. The pages used varied in colour depending on the level at which a particular title was displayed. These pages were all spiral bound once they had been assembled in the correct order. The paper model can be compared to an address diary where one can see the letters o f the alphabet upon opening the front

cover. When a particular letter is chosen, the pages reveal the addresses and

1 AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 49

In a similar way, the conceptual model of the AIDS Information System showed the top level menu headings of the General Public system as this would appear on the computer. When one o f these was chosen and the relevant page opened, a sub­ menu appeared reflecting the chosen item. This process continued until the bottom of the structure was reached, i.e. no more menus appear. This meant that a report was attached to that item. To restart, all the pages were flicked back to their original positions, enabling another choice to be made.

This model was sent to Dr. Farthing together with a printout of all the relevant reports. Using this model he was able to check the structure of the General Public system and the contents o f the reports which were associated with each item. He could use this model whenever he had the opportunity and was able to carry the compact conceptual model in his pocket.

The knowledge engineers noticed the benefits of sending this model on the very first day of their next visit to see the expert. He was already familiar with the new structure o f the General Public system and had also made comments and

corrections to some of the printouts that were sent out to him. Therefore,

invaluable time was saved and the knowledge engineers could concentrate on other modules and aspects of the system.

One o f the problems with a conceptual model is that it is difficult to keep up-to- date once the structure is changed. Either it has to be re-made completely or the pages removed, updated and modified before re-binding them. Both of these methods would be very time-consuming but, fortunately, the structure of the General Public module did not need changing very much after the first major change. Therefore, the model had already served its purpose. This concept of

1: AN EXPERT ADVISORY SYSTEM FOR AIDS/HIV 50

updating also highlights one of the advantages of having a computerised system as opposed to a paper-based one.

1.5.3.2 The Four Modules